Content Validity

According to Hair et al. (2006), validity is defined as when a scale or set of measures precisely represents the concept of interest. Previous researchers such as Davis (1989), Cavana, Delahaye and Sekaran (2001) argue that content validity is addressed by 1) reviewing relevant literature in order to identify suitable items; 2) seeking expert advice in order to sharpen the items; 3) conducting pilot tests in order to check the clarity of items and wording; and 4) relevant modification. Therefore, this study has followed these

95

steps, starting with the following. Firstly, the researcher reviewed the relevant literature in order to obtain a suitable and validated existed instrument that will be used in this research. Secondly, four PhD researchers were consulted to provide their judgment on the questionnaires and specifically the items in it. Thirdly, a pilot study was conducted by distributing fifty surveys 30 to employees and 20 to students, in order to check the clarity of items and the wording without modifying the original scale. Finally, some relevant modifications were made to the instrument according to the feedback in order to ensure the content validity of the instrument such as a reduction in the number of questions from 48 to 34 to remove questions that seemed to be vague, the addition of some definitions at the beginning about m-Government and Mobile Parking Service and the restatement of some questions that seemed to be repetitive to ensure greater readability and clarity.

4.7.2. Construct Validity

Construct validity is directly related to the questions that the instrument is measuring. There are two forms of construct validity that seem to be accepted widely, convergent, and discriminate validity (Churchill & Iacobucci 2009). The former is where indicators of a specific construct covers or share a high proportion of variance in common (Hair et al.2006). The latter is when the measures among different constructs have a low correlation (Zikmund 2003). This research ascertained the validity of the questions during the development of the instrument. In addition, the variables to be measured in this research were obtained from the well-known theories of Rogers’ Diffusion of Innovations (DOI) as well as the Technology Accepted Model (TAM) and others from existing scales derived from the literature. These variables have been measured by previous researchers and have been confirmed as having construct validity. Thus, questions that have been confirmed as valid by previous researchers have been adapted to suit this research and the wording amended as necessary without affecting the original scale.

96

4.7.3. Reliability

The research might produce inaccurate results and negate the acceptance or rejection of hypotheses if reliability and validity are not checked (Creswell 2009). Consequently, reliability and validity are two vital characteristics of questionnaire items in a good measurement instrument (Cavana, Delahaye, & Sekaran, 2001; Zikmund 2003; Cresswell 2009). Reliability is imperative but not an essential provision for validity (Cavana, Delahaye, & Sekaran, 2001). For instance, a test may not be valid but may be reliable. On the other hand, a test must be valid in order to be a reliable. According to Zikmund (2003), reliability is a measure where similar outcomes are achieved over time and across conditions. Pallant (2005) agrees that reliability tests the consistency and stability of a measurement instrument.

Litwin (1995) suggests that there are four different methods to measure reliability 1) test- retest; 2) intra-observer; 3) inter-observer and; 4) internal consistency. This research utilizes internal consistency, as it is more suitable than the others. In order to measure internal consistency or reliability, Cronbach’s alpha will be used, and the commonly accepted rule of thumb is that the result should be 0.8 and above (Litwin 1995; Malhotra & Birks 2000; & Bryman 2007). Chapter 5 provides more details about the results from Cronbach’s alpha.

4.7.4. External Validity

This study has adapted the seven staged sampling procedure developed by Zikmund (2003) as mentioned before. Table 4.8 describes the chosen sample of population in relation to Zikmund’s seven stages. The sample clearly reflects the general population in Oman. Therefore, outcomes reached in this study can be generalised to the rest of the population. Having said that, caution has to be taken as the sample and the mobile service (Mobile Parking) is specific to the capital area, Muscat. However, as the Omani context has no unique factors in terms of m-Government services, the model proposed in this research will be useful for many countries considering delivery of m-Government

97

services in order to explain the factors that influence/impact the intention to use m- Government services and it will contribute to the existing knowledge.

4.8.

Data Analysis

In the second semester of 2009, about 300 questionnaires were distributed and 253 responses were collected. Among these 253 responses, 7 responses were discarded due to large proportions of the questionnaires not being answered. Therefore, 246 responses were deemed to be usable which is about (82 per cent response rate).

4.8.1. Survey analysis

The research model introduced in the previous Chapter three consists of 9 independent variables and 1 dependent variable. These variables need to be analysed using an appropriate analysis tool. Therefore, for this study different tools were used for the data analysis. First, the Statistical Package for the Social Sciences (SPSS) version 16 was used to organize the data that had been collected. Second, all the data were divided into segments, i.e. males, females, young and old people, users of the Mobile Parking Service, non-users of the Mobile Parking Service, highly educated, and less educated people. These segments were then transformed into Excel spreadsheet. However, for testing the model and hypotheses, the advanced technique of Structural Equation Modeling (SEM) was used (see Chapter 5). For analysing results of variables in particular latent constructs that have multiple dimensions in order to allow assessment of measurement properties and theoretical (structural) relationship, SEM is considered a powerful second-generation multivariate technique to meet this purpose (Hoyle 1995; Kline 2005; & Maruyama 1997). This study adopts Partial Least Squares (PLS) path modeling which is a division of SEM, and widely employed in information systems and marketing research. In addition, this PLS path modeling was chosen it is considered more appropriate for predictive and exploratory research (Azwadi, 2010). Further, PLS supports both formative and reflective variables (Thompson, Barclay, & Higgins, 1995; Chin, 1998) and can support exploratory and confirmatory research, whereas, other

98

methods as LISERL and Amos can only support reflective constructs (Gefen, Straub, & Boudreau, 2000).

The software used for PLS path modeling is called SmartPLS 2.0 tool developed by (Ringle, Wende & Will 2005). According to Bagozzi and Fornell (1982), the second- generation data analysis technique of SEM provides more benefits than the traditional first generation statistical tools such as ANOVA and MANOVA. In addition, Gefen, Straub, and Boudreau, (2000) argue that using the SEM technique allows the researcher to run a systematic and comprehensive analysis in order to test the interconnected variables and their items with only one single run, whereas the first generation tools can only analyze one layer of linkages between variables at a time. Further, there are four advantages of using the SEM technique as proposed by Bryman (2006). They are as follows:

 SEM takes mainly a confirmatory approach rather than an exploratory approach to the data analysis;

 SEM provides clear estimates of errors variance parameters;

 Data analysis using SEM procedures determines both unobserved (e.g. latent) as well as observed variables; and

 SEM methodology has several crucial features such as modeling multivariate relations, and estimating point and/or interval indirect effects.

Because of these advantages the SEM, technique was considered the most appropriate for this research in order to test the research model against the data. Using SEM method, the data analysis for the survey questionnaires go through several steps. These are as follows (for more details see Chapter 5):

 Distribution of Latent Variables: The aim is to present a broad picture of the respondent’s evaluation of each perception or variable in the research model and understand more about the characteristics of the sample;

 Construct Specification Accuracy: This is to distinguish between the relationship of measures and latent constructs in order to avoid misspecification;

99

 Measurement model: is used to validate the indicators that are used to measure the latent variables using a confirmatory factor analysis. In order to specify a valid measurement model, it is imperative to establish satisfactory convergent and discriminant validities for the research model. The former is done through establishing Reliabilities of items in relation to their constructs, composite reliabilities (CR) of constructs, and the average variance extracted (AVE) are used in order to assess the convergent validity, whereas the latter is established for two reasons. Firstly, when all the items that are used to measure a construct load highly on that construct compared to their loadings on other constructs in the research model;

 Structural Model: is established when the R square value in a structural equation model measures the amount of variance in the dependent variable that an independent variable explains. As a rule of thumb, this R square value for endogenous variables should be higher or equal to 0.10; and

 Confirmation of Hypotheses: Each link in the structural model represents a hypothesis to be tested and for this research, there are ten hypotheses to be examined. Testing the hypothesis via Partial Least Squares (PLS) using Structural Equation Modeling (SEM) occurs through the calculation of the strength and the significance/insignificance of every structural path which in turn indicate that the hypothesis is confirm or otherwise.

4.8.2. Interview analysis

On the other hand, the software program used for analysis in this research for qualitative interviews was the NVivo computer program. The software program used for analysis in this research for qualitative interviews was the NVivo computer program. This program has several benefits; for example, it can easily manage, access and analyse qualitative data without losing its richness. Further, it is practical for locating patterns or common threads, and can be used to develop greater or more subtle concepts (Bazeley & Richards, 2000). According to Cavana, Delahaye, and Sekaran (2001), the NVivo program is one of the most popular programs available, and is well suited for analysis of

100

both unstructured and semi- structured interviews in qualitative research. Furthermore, this program is practical for small numbers of interviews where discussions are recorded. As mentioned in Section 4.6, the decision makers involved with m-Government projects were interviewed using a semi-structured format. The decision makers were asked whether they wanted to have the interview recorded on tape or whether they preferred the researcher to take notes. In addition, they were asked whether they preferred to use Arabic or English during the interviews. All of them agreed to recorded interviews in English.

Figure 4.3 shows the data analysis approach used in the qualitative method adapted from Creswell (2009) followed in this research building from bottom to top.

Step 1. Organizing and preparing the data for analysis. This involves transcribing the semi-structured interviews;

Step 2. Reading all data in order to obtain a general sense of data as well as an understanding of the general ideas of the participants' views;

Step 3. Coding of the data. This step involves organizing and categorizing all data into segments;

Step 4. Generating themes for all the data to prepare for the analysis;

Step 5. Interrelating all themes in order to provide a discussion of themes in the qualitative narrative; and

Step 6. Interpreting the meaning of themes, and learning from the data analysis process.

101

Coding the Data

Interpreting the Meaning of Themes

Themes

Raw Data (Transcripts) Reading Through All Data

Interrelating Themes

Organizing and Preparing for Data Analysis Validating the Accuracy of

the Information

Figure 4.3 The data analysis approach used in the qualitative method

Source: adapted from Creswell (2009)

The coding process was done using the guidelines recommended by Tesch (1990). These guidelines are as follows:

 The researcher carefully read all the transcriptions in order to get a sense of the whole data;

 The researcher went through the first interview. This was the most interesting interview because this participant was fully involved with the Mobile

Parking Service (the example of m-Government services used in this research). Then, several questions were noted; for instance, ‘What is this about?’ and ‘What does this mean?’;

102

 After completing the above task for the remaining participants, a list of themes was made and these themes were then grouped into similar topics.

 Those topics were abbreviated as codes and were written next to the appropriate segments of the text;

 Those topics were then turned into categories by grouping the topics that related to each other;

 The final decision on the appropriate abbreviation for each category was made; and

 The data was then ready for the preliminary analysis.